If you donât mind me expanding on #5, bc as someone who does a LOT of research based projects for physics and psychology (which are primarily statistics and probability) this report bothers me SO much:
âThe reader should also note that the data presented are extensively corrected for the existence of any bias.â This claim is not properly explained within the report or the announcements made by the moderators. There is no clarification on how bias was âcorrected.â And even then, bias is not âcorrectedâ in any research. All data is inherently biased (an unavoidable fact that shows that the people running this investigation donât understand statistics or research on a fundamental level). As well as that, personal biases towards the individual being investigated should not influence the research into what is meant to be a formal investigation at all. If someone is biased in this, they should not partake in the investigation. Plain and simple.
The Data section consists of only two sets of data, each accounting for only two observed data sets and two predicted sets. This is not enough data to make any substantial claims, especially given that the team spent two months on this. While extra data is included in the appendices, this is incredibly improper formatting given the data presented, and leads to the larger set of data not being included in the official findings. (the way that my physics prof wouldâve fucking strangled me if i tried that shit i swear to god)
The pre-emptive FAQ in the analysis does not belong anywhere in this report. If you want a statistical report, then this informal discussion needs to be posted separately.Â
âYes. There is clearly sampling bias in the data set, but its presence does not invalidate our analysis.â No, the bias does not impact the data collected, but it does impact the analysis.
âSampling bias is a common problem in real-world statistical analysis, so if it were impossible to account for, then every analysis of empirical data would be biased and useless.â In real-world statistical analysis, researches will use as large of a sample as is realistic and possible in an attempt to eliminate as much sample bias as possible. They donât say âoh well what can you doâ and sample only six trials. When there is insufficient data, they will make this clear. Remember when scientists found particles that moved reverse of how they were expected to earlier this year? They didnât start claiming it to be proof of a parallel universe (although media did), they explained that there were countless mundane explanations and simply kept the data on record, knowing that there was nowhere near enough information to present a guess as a factual result. Also, the moderators comparing this investigation to actual formal research creates a false sense that this investigation is comparable to âreal-world statistical analysisâ when itâs not.
âHowever, what luck Dream actually got in any other instance is irrelevant to this analysis, as it has absolutely no bearing in how likely the luck was in this instance.â A researcher cannot pick and choose which data they use like this. Other trials were painfully needed in this investigation, even if viewed separately from the stream in question, they could serve as a basis of expected behavior worth comparing. Again, two months were spent on this, and it does not show. (sidenote: donât use italics for emphasis in formal writing, extremely unprofessional.)
There is no replication to serve as a sample set of data. Calculating the expected results and using other peopleâs streams is not a reliable basis, actual trials should be run by the research team themselves.
The entire inclusion of âShifty Samâ is beyond unprofessional. Like the FAQ, it should have been used separately from the findings themselves, if at all.
Failure to properly acknowledge the fact that a speedrunner who is experiencing good luck will continue to play longer, which can create a false sense of an increase in luck, as was the case in the stream in question.
The data compares a lucky stream with average luck streams of other speedrunners, of course one is going to seem statistically different from the rest. Comparing the lucky stream in question with other lucky streams or record setting runs would have potentially yielded more precise and accurate results.Â
âThis is 1 in 20 sextillion. The idea that something this unlikely occurred is obviously ridiculous.â No, the idea that something that unlikely occurring is actually 1 in 20 sextillion, this is so unprofessional it isnât even funny.
The entire section combining pearl trades and blaze drops is irrelevant to the investigation at hand, and is done solely to provide examples of decreased probability that make the rest seem unlikely as well.
No discussion of how bias was avoided in data collection, handling of the investigation, presentation of findings, etc.
âthe likelihood of [the pearl trades] occurring is still unfathomably small. There are no circumstances in a natural setting in which bartering and blaze drops could be dependent or biased to any notable degree, much less a degree strong enough to produce this result.â The data found and research provided quite literally shows that, while unlikely, it is possible for these events to occur naturally.
Minecraft speedrunning comes down to luck, until youâre too lucky, apparently.
